The world is one big dataset. Now, how to photograph it ...
-
0:00 - 0:03Five years ago, I was a Ph.D. student
-
0:03 - 0:05living two lives.
-
0:05 - 0:07In one, I used NASA supercomputers
-
0:07 - 0:10to design next-generation spacecraft,
-
0:10 - 0:12and in the other I was a data scientist
-
0:12 - 0:15looking for potential smugglers
-
0:15 - 0:18of sensitive nuclear technologies.
-
0:18 - 0:21As a data scientist, I did a lot of analyses,
-
0:21 - 0:22mostly of facilities,
-
0:22 - 0:25industrial facilities around the world.
-
0:25 - 0:27And I was always looking for a better canvas
-
0:27 - 0:29to tie these all together.
-
0:29 - 0:31And one day, I was thinking about how
-
0:31 - 0:33all data has a location,
-
0:33 - 0:35and I realized that the answer
-
0:35 - 0:37had been staring me in the face.
-
0:37 - 0:40Although I was a satellite engineer,
-
0:40 - 0:43I hadn't thought about using satellite imagery
-
0:43 - 0:44in my work.
-
0:44 - 0:46Now, like most of us, I'd been online,
-
0:46 - 0:48I'd see my house, so I thought,
-
0:48 - 0:50I'll hop in there and I'll start looking up
-
0:50 - 0:52some of these facilities.
-
0:52 - 0:54And what I found really surprised me.
-
0:54 - 0:55The pictures that I was finding
-
0:55 - 0:57were years out of date,
-
0:57 - 0:58and because of that,
-
0:58 - 1:00it had relatively little relevance
-
1:00 - 1:03to the work that I was doing today.
-
1:03 - 1:04But I was intrigued.
-
1:04 - 1:07I mean, satellite imagery is pretty amazing stuff.
-
1:07 - 1:10There are millions and millions of sensors
-
1:10 - 1:11surrounding us today,
-
1:11 - 1:14but there's still so much we
don't know on a daily basis. -
1:14 - 1:18How much oil is stored in all of China?
-
1:18 - 1:21How much corn is being produced?
-
1:21 - 1:25How many ships are in all of our world's ports?
-
1:25 - 1:27Now, in theory, all of these questions
-
1:27 - 1:30could be answered by imagery,
-
1:30 - 1:31but not if it's old.
-
1:31 - 1:34And if this data was so valuable,
-
1:34 - 1:36then how come I couldn't get my hands
-
1:36 - 1:38on more recent pictures?
-
1:38 - 1:41So the story begins over 50 years ago
-
1:41 - 1:43with the launch of the first generation
-
1:43 - 1:47of U.S. government photo reconnaissance satellites.
-
1:47 - 1:48And today, there's a handful
-
1:48 - 1:51of the great, great grandchildren
-
1:51 - 1:52of these early Cold War machines
-
1:52 - 1:54which are now operated by private companies
-
1:54 - 1:57and from which the vast majority of satellite imagery
-
1:57 - 2:00that you and I see on a daily basis comes.
-
2:00 - 2:03During this period, launching things into space,
-
2:03 - 2:05just the rocket to get the satellite up there,
-
2:05 - 2:10has cost hundreds of millions of dollars each,
-
2:10 - 2:12and that's created tremendous pressure
-
2:12 - 2:14to launch things infrequently
-
2:14 - 2:15and to make sure that when you do,
-
2:15 - 2:19you cram as much functionality in there as possible.
-
2:19 - 2:21All of this has only made satellites
-
2:21 - 2:23bigger and bigger and bigger
-
2:23 - 2:25and more expensive,
-
2:25 - 2:30now nearly a billion, with a b, dollars per copy.
-
2:30 - 2:31Because they are so expensive,
-
2:31 - 2:33there aren't very many of them.
-
2:33 - 2:34Because there aren't very many of them,
-
2:34 - 2:37the pictures that we see on a daily basis
-
2:37 - 2:38tend to be old.
-
2:38 - 2:42I think a lot of people actually
understand this anecdotally, -
2:42 - 2:44but in order to visualize just how sparsely
-
2:44 - 2:46our planet is collected,
-
2:46 - 2:48some friends and I put together a dataset
-
2:48 - 2:51of the 30 million pictures that have been gathered
-
2:51 - 2:54by these satellites between 2000 and 2010.
-
2:54 - 2:57As you can see in blue, huge areas of our world
-
2:57 - 3:00are barely seen, less than once a year,
-
3:00 - 3:02and even the areas that are seen most frequently,
-
3:02 - 3:06those in red, are seen at best once a quarter.
-
3:06 - 3:09Now as aerospace engineering grad students,
-
3:09 - 3:12this chart cried out to us as a challenge.
-
3:12 - 3:15Why do these things have to be so expensive?
-
3:15 - 3:18Does a single satellite really have to cost
-
3:18 - 3:23the equivalent of three 747 jumbo jets?
-
3:23 - 3:25Wasn't there a way to build a smaller,
-
3:25 - 3:28simpler, new satellite design that could enable
-
3:28 - 3:30more timely imaging?
-
3:30 - 3:34I realize that it does sound a little bit crazy
-
3:34 - 3:35that we were going to go out and just
-
3:35 - 3:37begin designing satellites,
-
3:37 - 3:39but fortunately we had help.
-
3:39 - 3:42In the late 1990s, a couple of professors
-
3:42 - 3:45proposed a concept for radically reducing the price
-
3:45 - 3:47of putting things in space.
-
3:47 - 3:49This was hitchhiking small satellites
-
3:49 - 3:52alongside much larger satellites.
-
3:52 - 3:55This dropped the cost of putting objects up there
-
3:55 - 3:57by over a factor of 100,
-
3:57 - 4:00and suddenly we could afford to experiment,
-
4:00 - 4:02to take a little bit of risk,
-
4:02 - 4:04and to realize a lot of innovation.
-
4:04 - 4:07And a new generation of engineers and scientists,
-
4:07 - 4:09mostly out of universities,
-
4:09 - 4:11began launching these very small,
-
4:11 - 4:14breadbox-sized satellites called CubeSats.
-
4:14 - 4:16And these were built with electronics obtained
-
4:16 - 4:20from RadioShack instead of Lockheed Martin.
-
4:20 - 4:23Now it was using the lessons
learned from these early missions -
4:23 - 4:26that my friends and I began a series of sketches
-
4:26 - 4:27of our own satellite design.
-
4:27 - 4:30And I can't remember a specific day
-
4:30 - 4:32where we made a conscious decision
-
4:32 - 4:35that we were actually going to
go out and build these things, -
4:35 - 4:37but once we got that idea in our minds
-
4:37 - 4:39of the world as a dataset,
-
4:39 - 4:42of being able to capture millions of data points
-
4:42 - 4:45on a daily basis describing the global economy,
-
4:45 - 4:47of being able to unearth billions of connections
-
4:47 - 4:50between them that had never before been found,
-
4:50 - 4:52it just seemed boring
-
4:52 - 4:55to go work on anything else.
-
4:55 - 4:58And so we moved into a cramped,
-
4:58 - 5:01windowless office in Palo Alto,
-
5:01 - 5:03and began working to take our design
-
5:03 - 5:06from the drawing board into the lab.
-
5:06 - 5:08The first major question we had to tackle
-
5:08 - 5:11was just how big to build this thing.
-
5:11 - 5:14In space, size drives cost,
-
5:14 - 5:16and we had worked with these very small,
-
5:16 - 5:18breadbox-sized satellites in school,
-
5:18 - 5:21but as we began to better
understand the laws of physics, -
5:21 - 5:23we found that the quality of pictures
-
5:23 - 5:26those satellites could take was very limited,
-
5:26 - 5:28because the laws of physics dictate
-
5:28 - 5:31that the best picture you
can take through a telescope -
5:31 - 5:33is a function of the diameter of that telescope,
-
5:33 - 5:35and these satellites had a very small,
-
5:35 - 5:36very constrained volume.
-
5:36 - 5:38And we found that the best picture we would
-
5:38 - 5:40have been able to get looked something like this.
-
5:40 - 5:43Although this was the low-cost option,
-
5:43 - 5:44quite frankly it was just too blurry
-
5:44 - 5:48to see the things that make
satellite imagery valuable. -
5:48 - 5:50So about three or four weeks later,
-
5:50 - 5:53we met a group of engineers randomly
-
5:53 - 5:55who had worked on the first
-
5:55 - 5:57private imaging satellite ever developed,
-
5:57 - 5:59and they told us that back in the 1970s,
-
5:59 - 6:01the U.S. government had found a powerful
-
6:01 - 6:03optimal tradeoff --
-
6:03 - 6:06that in taking pictures at right
about one meter resolution, -
6:06 - 6:09being able to see objects one meter in size,
-
6:09 - 6:12they had found that they could not
just get very high-quality images, -
6:12 - 6:15but get a lot of them at an affordable price.
-
6:15 - 6:17From our own computer simulations,
-
6:17 - 6:19we quickly found that one meter really was
-
6:19 - 6:20the minimum viable product
-
6:20 - 6:23to be able to see the drivers of our global economy,
-
6:23 - 6:25for the first time, being able to count
-
6:25 - 6:28the ships and cars and shipping
containers and trucks -
6:28 - 6:30that move around our world on a daily basis,
-
6:30 - 6:34while conveniently still not
being able to see individuals. -
6:34 - 6:36We had found our compromise.
-
6:36 - 6:37We would have to build something larger
-
6:37 - 6:39than the original breadbox,
-
6:39 - 6:41now more like a mini-fridge,
-
6:41 - 6:43but we still wouldn't have to build a pickup truck.
-
6:43 - 6:46So now we had our constraint.
-
6:46 - 6:48The laws of physics dictated
-
6:48 - 6:51the absolute minimum-sized
telescope that we could build. -
6:51 - 6:54What came next was making the rest of the satellite
-
6:54 - 6:56as small and as simple as possible,
-
6:56 - 6:59basically a flying telescope with four walls
-
6:59 - 7:02and a set of electronics smaller than a phone book
-
7:02 - 7:05that used less power than a 100 watt lightbulb.
-
7:05 - 7:07The big challenge became actually taking
-
7:07 - 7:09the pictures through that telescope.
-
7:09 - 7:12Traditional imaging satellites use a line scanner,
-
7:12 - 7:14similar to a Xerox machine,
-
7:14 - 7:16and as they traverse the Earth, they take pictures,
-
7:16 - 7:18scanning row by row by row
-
7:18 - 7:20to build the complete image.
-
7:20 - 7:23Now people use these because they get a lot of light,
-
7:23 - 7:25which means less of the noise you see
-
7:25 - 7:28in a low-cost cell phone image.
-
7:28 - 7:30The problem with them is they require
-
7:30 - 7:32very sophisticated pointing.
-
7:32 - 7:35You have to stay focused on a 50-centimeter target
-
7:35 - 7:37from over 600 miles away
-
7:37 - 7:39while moving at more than
seven kilometers a second, -
7:39 - 7:42which requires an awesome degree of complexity.
-
7:42 - 7:45So instead, we turned to a new
generation of video sensors, -
7:45 - 7:48originally created for use in night vision goggles.
-
7:48 - 7:51Instead of taking a single, high quality image,
-
7:51 - 7:52we could take a videostream
-
7:52 - 7:55of individually noisier frames,
-
7:55 - 7:57but then we could recombine
-
7:57 - 7:59all of those frames together
-
7:59 - 8:01into very high-quality images
-
8:01 - 8:03using sophisticated pixel processing techniques
-
8:03 - 8:05here on the ground,
-
8:05 - 8:08at a cost of one one hundredth a traditional system.
-
8:08 - 8:09And we applied this technique
-
8:09 - 8:12to many of the other systems on the satellite as well,
-
8:12 - 8:15and day by day, our design evolved
-
8:15 - 8:18from CAD to prototypes
-
8:18 - 8:21to production units.
-
8:21 - 8:23A few short weeks ago,
-
8:23 - 8:25we packed up SkySat 1,
-
8:25 - 8:26put our signatures on it,
-
8:26 - 8:29and waved goodbye for the last time on Earth.
-
8:29 - 8:32Today, it's sitting in its final launch configuration
-
8:32 - 8:35ready to blast off in a few short weeks.
-
8:35 - 8:38And soon, we'll turn our attention to launching
-
8:38 - 8:41a constellation of 24 or more of these satellites
-
8:41 - 8:43and beginning to build the scalable analytics
-
8:43 - 8:45that will allow us to unearth the insights
-
8:45 - 8:49in the petabytes of data we will collect.
-
8:49 - 8:53So why do all of this? Why build these satellites?
-
8:53 - 8:55Well, it turns out imaging satellites
-
8:55 - 8:59have a unique ability to provide global transparency,
-
8:59 - 9:02and providing that transparency on a timely basis
-
9:02 - 9:05is simply an idea whose time has come.
-
9:05 - 9:09We see ourselves as pioneers of a new frontier,
-
9:09 - 9:10and beyond economic data,
-
9:10 - 9:14unlocking the human story, moment by moment.
-
9:14 - 9:15For a data scientist
-
9:15 - 9:18that just happened to go to space camp as a kid,
-
9:18 - 9:21it just doesn't get much better than that.
-
9:21 - 9:23Thank you.
-
9:23 - 9:27(Applause)
- Title:
- The world is one big dataset. Now, how to photograph it ...
- Speaker:
- Dan Berkenstock
- Description:
-
We're all familiar with satellite imagery, but what we might not know is that much of it is out of date. That's because satellites are big and expensive, so there aren't that many of them up in space. As he explains in this fascinating talk, Dan Berkenstock and his team came up with a different solution, designing a cheap, lightweight satellite with a radically new approach to photographing what's going on on Earth.
- Video Language:
- English
- Team:
- closed TED
- Project:
- TEDTalks
- Duration:
- 09:44
janet dragojevic edited English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Morton Bast approved English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Morton Bast edited English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Morton Bast edited English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Madeleine Aronson edited English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Madeleine Aronson accepted English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Joseph Geni edited English subtitles for The world is one big dataset. Now, how to photograph it ... | ||
Amara Bot edited English subtitles for The world is one big dataset. Now, how to photograph it ... |